The main drivers of arabica coffee prices in Latin America - PROJECT DOCUMENTS Javier Aliaga Lordemann Claudio Mora-García Nanno Mulder

Page created by Randy Sanders
 
CONTINUE READING
The main drivers of arabica coffee prices in Latin America - PROJECT DOCUMENTS Javier Aliaga Lordemann Claudio Mora-García Nanno Mulder
PROJECT DOCUMENTS

The main drivers of arabica
coffee prices in Latin America
Javier Aliaga Lordemann
Claudio Mora-García
Nanno Mulder
Thank you for your interest in
 this ECLAC publication

 ECLAC
 Publications

 Please register if you would like to receive information on our editorial
products and activities. When you register, you may specify your particular
areas of interest and you will gain access to our products in other formats.

 Register

 www.cepal.org/en/publications

 facebook.com/publicacionesdelacepal

 Publicaciones
 www.cepal.org/apps
Project Documents

The main drivers of arabica coffee prices
 in Latin America
 Javier Aliaga Lordemann
 Claudio Mora‐García
 Nanno Mulder
This document was prepared by Javier Aliaga Lordemann, Chief of the Regional Climate Change Program of the
Latin American and Caribbean Network of Fair Trade Small Producers and Workers (CLAC); Claudio A. Mora‐García,
Professor at the University of Costa Rica and Senior Researcher at INCAE Business School, in Costa Rica; and Nanno
Mulder, Chief of the International Trade Unit of the Division of Internation trade and Integration of the Economic
Commission for Latin America and the Caribbean (ECLAC). It was prepared under the 2020 cooperation agreement
between CLAC and ECLAC. The authors are grateful for the comments received from the participants in a workshop
held on 3 November, 2020.
The views expressed in this document, which has been reproduced without formal editing, are those of the authors
and do not necessarily reflect the views of the Organization.

United Nations publication
LC/TS.2021/25
Distribution: L
Copyright © United Nations, 2021
All rights reserved
Printed at United Nations, Santiago
S.21‐00105

This publication should be cited as: J. Aliaga Lordemann, C. Mora‐García and N. Mulder, “The main drivers of arabica coffee
prices in Latin America”, Project Documents, (LC/TS.2021/25), Santiago, Economic Commission for Latin America and the
Caribbean (ECLAC), 2021.

Applications for authorization to reproduce this work in whole or in part should be sent to the Economic Commission for
Latin America and the Caribbean (ECLAC), Documents and Publications Division, publicaciones.cepal@un.org. Member States and
their governmental institutions may reproduce this work without prior authorization, but are requested to mention the source and
to inform ECLAC of such reproduction.
ECLAC The main drivers of arabica coffee prices in Latin America 3

 Contents

Abstract................................................................................................................................................ 5
Introduction ......................................................................................................................................... 7
I. Literature review ...................................................................................................................... 9
II. A simple pricing model for arabica coffee ............................................................................... 15
III. Definition of variables and data sources ................................................................................. 17
 A. Data sources for factors that shift the supply curve ........................................................... 17
 B. Data sources for factors that shift the demand curve ........................................................ 19
 C. Stocks ............................................................................................................................... 19
 D. Prices ................................................................................................................................ 19
 E. Data description ................................................................................................................ 20
IV. Methods and results ................................................................................................................ 25
 A. Model 1: ICO composite prices (monthly, Jan‐1997 to Sep‐2020) ...................................... 25
 B. Model 2: colombian prices and production (monthly, Jan‐1997 to Jun‐2020) ..................... 27
 C. Model 3: panel data for farm gate prices (annual, panel data, 1979 to 2019) ......................29
V. Discussion and conclusions ..................................................................................................... 33
Bibliography ....................................................................................................................................... 35
Annex ................................................................................................................................................. 37
ECLAC The main drivers of arabica coffee prices in Latin America 4

Tables
Table 1 Data sources and definitions of factors that shift the supply curve
 and its data sources ................................................................................................... 18
Table 2 Data sources and definitions of factors that shift the demand curve
 and its data sources ................................................................................................... 19
Table 3 Data sources and definitions of stocks prices ............................................................. 20
Table 4 Decade average of green coffee yields in coffee growing Latin American
 countries, 1961‐2018 ................................................................................................. 22
Table 5 Determinants of monthly ICO prices from Arabica coffee (our Arabica category,
 Brazilian Naturals, Colombian Naturals, and Other Milds) using Model 1 ..................26
Table 6 Determinants of the monthly prices of colombian coffee (Colombian Milds, price
 paid to Colombian coffee growers) and quantity produced using Model 2 ................28
Table 7 Determinants of annual farm gate prices using panel data for Latin American
 countries, 1979‐2019 and Model 3.............................................................................. 30
Table A1 Summary of the literature, methods and significant variables affecting
 coffee prices .............................................................................................................. 39
Table A2 Variance Inflation Factors (VIF) and Dickey‐Fuller statistics
 for the variables in Model 1 ........................................................................................ 41
Table A3 Variance Inflation Factors (VIF) and Dickey‐Fuller statistics
 for the variables in Model 2 ........................................................................................ 41
Table A4 Variance Inflation Factors (VIF) for the variables in Model 3 ....................................... 41
Table A5 Classification of exporter countries by type of coffee according to ICO......................42

Figures
Figure 1 Arabica coffees: weekly spot prices for coffee ‘C’ contract traded
 at ICE and ICO composite prices for the three categories (Brazil,
 Colombia and other milds), 1997‐2020 ...................................................................... 20
Figure 2 Arabica coffees: weekly price differentials between Coffee ‘C’ Contract spot
 prices and ICO Composite prices for the three categories, 1997‐2020 ....................... 21
Figure 3 Annual yield (hg/ha) of green coffee in Latin American countries, 1961‐2018 ............ 23
Figure 4 Monthly relationship between the real exchange rate of the Brazilian Real
 to the US Dollar and the real spot price, Jan 1994‐Jun 2020 .......................................24
Figure 5 The effect of temperature and rain on farm gate prices using Model 3....................... 31
Figure 6 Heterogeneity in the pass‐through coefficients from international spot prices
 to national farm gate prices using Model 3 ................................................................ 32
Figure A1 Arabica green coffee production and exports (in 1,000 60 kg bags)
 by Latin American and Caribbean (LAC) countries during the 2018/2019 harvest ...... 38

Diagrams
Diagram 1 Factors affecting agricultural commodity prices ........................................................ 11
Diagram 2 Conceptual framework for factors contributing to coffee production costs ............... 12
ECLAC The main drivers of arabica coffee prices in Latin America 5

 Abstract

This paper analyzes the determinants of arabica green coffee prices in Latin American countries using a
time series analysis and panel data methods. For this purpose, we construct a panel of different coffee
prices: Coffee Organization (ICO) composite prices for Brazilian Naturals, Colombian Milds, and Other
Milds; prices paid by the Federación Nacional de Cafeteros de Colombia (FNC) to coffee growers; and farm
gate prices by country. The results show that the Brazilian Real to USD real exchange rate, inflation, and
rain in January affect prices positively. In contrast, green coffee inventories, the oil price, and the
Colombian Peso to USD real exchange rate negatively affect coffee prices.
ECLAC The main drivers of arabica coffee prices in Latin America 7

 Introduction

In this paper, we study the determinants of Arabica green coffee prices in Latin America and Caribbean
(LAC) using a rigorous econometric analysis. The topic is of great practical importance, as green coffee
is one of the leading export products of many Latin American and Caribbean countries. This commodity
is produced by hundreds of thousands of families in the region, mostly smallholder farmers. The study
of the determinants of Arabica green coffee prices is essential for understanding and predicting coffee
prices and the improvement of economic policymaking in these countries (Deaton and Laroque, 1992).
Arabica, together with Robusta, are the two predominant coffee species produced by more than
50 countries worldwide inside the "Coffee Belt", i.e., the region located between 25°N and 30°S,
between the Tropic of Cancer and the Tropic of Capricorn.
 Our analysis focuses on Arabica coffee prices instead of Robusta prices in LAC for several reasons.
First, Arabica and Robusta prices may have different drivers: each caters to different market segments;1 the
former is more susceptible to diseases than the latter: and each requires other farming techniques and use
different inputs. In 2020, the global Arabica coffee production was 2.4 times the Robusta production (United
States Department of Agriculture, USDA). Second, in 2020 Latin America accounted for 84% of the global
Arabica green coffee production and 51% of the global Robusta green coffee production.
 We follow a time series methodology to assess the importance of each of the determinants in
explaining Arabica green coffee prices. We consider several prices at different levels of the supply chain:
International Coffee Organization (ICO) composite prices for Brazilian Naturals, Colombian Milds, and
Other Milds; national‐level annual farm gate prices collected by ICO, a panel dataset; and prices paid by
the National Federation of Coffee Producers (Federación Nacional de Cafeteros de Colombia, FNC) to
coffee growers. We consider a comprehensive set of determinants that may shift the supply and
demand curves and study how these factors influence the price in a reduced form setup. The complete
set of covariates include weather shocks and national and international macroeconomic variables.

1
 Their cup roast profile is different: Arabicas are more balanced and Robustas more stringent. Their use profile also differs: Robustas
 are used in espressos while Arabicas are used for regular brewed and specialty coffees.
ECLAC The main drivers of arabica coffee prices in Latin America 8

 Three different econometric models are estimated: two time‐series and one panel‐data. The first
analyzes the determinants of ICO composite prices using monthly observations from January 1997
to September 2020. The second looks into the drivers of the prices of Colombian Milds reported by ICO,
those paid to coffee growers by the FNC, and the amount of coffee produced in that country from
January 1997 to September 2020. The last model uses panel data to assess the determinants of farm
gate prices in Latin America collected by ICO.
 Our main findings from the ICO Composite model are:
 (i) The BRL/USD real exchange rate has a positive and statistically significant effect on all ICO
 prices. This effect is more prominent for Brazil Naturals;
 (ii) The end‐of‐month stocks both have a negative and statistically significant effect on all four
 ICO prices considered;
 (iii) An increase in other commodity prices have a positive and statistically significant effect on
 the prices of Brazilian Naturals and Other Milds, but it does not affect the Colombia ICO
 prices;
 (iv) There is no statistical evidence that the COP/USD real exchange rate, the international
 interest rate, nor the oil price affects ICO prices;
 (v) There is no evidence that national shocks (except for the interest rate) affect Colombian Mild
 ICO prices.
 Our main findings from the Colombian model are:
 (i) The Colombian Milds international price has a positive and statistically significant effect on
 the price paid by FNC to coffee producers;
 (ii) The WTI oil price and the COP/USD real exchange rate have a positive and statistically
 significant effect on prices paid to Colombian coffee growers by the FNC;
 (iii) We do not find statistical evidence that other variables affect Colombian coffee growers'
 prices. These include national inflation, national interest rate, inventories growth, the
 BRL/USD real exchange rate, the international interest rate, and coffee production in
 Colombia.
 Our main findings from the farm gate model using panel data are:
 (i) The GDP per capita growth rate has a positive and statistically significant effect on farm gate
 prices;
 (ii) National inflation has a positive and statistically significant effect on farm gate prices;
 (iii) Several other variables do not seem statistically significant, including domestic consumption,
 stocks at the end of the year, exports nor production affects farm gate prices;
 (iv) Rainfall in March and August has a significant negative effect, but rain in January has a
 significant positive impact on farm gate prices. Still, evidence for temperature is too noisy to
 draw any meaningful conclusion.
 The following section presents an extensive literature review to identify agricultural
commodities' main price drivers, including coffee. In section II, we describe the variables used in our
study. In turn, data sources, their mean features and empirical relationships are discussed in section III.
The methodology and regression results are presented in section IV. Finally, the last section concludes.
ECLAC The main drivers of arabica coffee prices in Latin America 9

 I. Literature review

Market prices result from the interaction between demand and supply and other macroeconomic and
financial variables. Hence, factors that shift the supply and demand curves will determine the
equilibrium price. The agricultural commodity price depends on market forces that alter the current or
expected balance between supply and demand. Initial models (Van Meir, 1983; and Baker and Menzie,
1988) did not formulate the price formation process and used reduced‐form expressions for prices
through a stocks‐to‐use ratio. They suggest this ratio explains price variations as stocks adjust to shocks
to supply and demand. Stocks will increase when production is high and prices are low, as sellers may
hold on entering in new transactions waiting for a higher price and execute a new sell. Total use includes
domestic consumption and exports, is generally more stable, and tends to shift gradually.
 Van Meir (1983) and Baker and Menzie (1988) analyzed the relationship between stocks‐to‐use
ratios and corn prices using annual data. Westcott, Hull, and Green (1984, 1985) used this approach in
models for quarterly wheat and corn prices. Westcott and Hoffman (1999) study the determinants of
farm‐level prices of wheat and corn and consider as determinants: (i) the ratio of total year‐end stocks
to use, (ii) the percentage of publicly held stocks owned by the Government (CCC) to use, and (iii) the
loan rate. Their wheat equation also included feed use and corn prices in the summer months. Their
empirical results confirmed a strong inverse relationship between price and the stocks‐to‐use ratio.
However, Goodwin, Schnepf, and Dohlman (2005) argue that the proportion of stocks to use is
endogenous to prices in a study that analyze the farm gate price of soybean.
 Subsequent studies provided a deeper understanding of the different factors affecting supply and
demand beyond the stocks to use ratio. These factors include consumer preferences and the changing
needs of end‐users; factors affecting production processes (e.g., weather, input costs, pests, diseases,
etc.); relative prices of crops that can substitute in either production or consumption; government
policies; and factors affecting storage and transportation (Schnepf, 2006).
 According to Tomek and Kaiser (2014), factors that shift the demand curve are those that affect
consumer preferences and budgets:
ECLAC The main drivers of arabica coffee prices in Latin America 10

 (i) Demographic factors: population size, age distribution, ethnicity, gender, etc.;
 (ii) Economic factors: income and its distribution, and the prices and availability of other
 products;
 (iii) Consumer tastes and preferences depend on education levels, life experiences, information
 and advertising, and the social context in which consumers live (i.e., lifestyle effects).
 Factors shifting the supply curve are those that affect marginal costs and marginal revenues are:
 (i) Changes in input prices;
 (ii) Changes in prices of commodities competing for the same resources or factors of
 production;
 (iii) Changes in the prices of joint products;
 (iv) Changes in the level of price faced by producers;
 (v) Changes in technology that influence efficiency and the costs of production; and
 (vi) Changes in institutional factors like government programs.
 Cooke and Robles (2009) study the determinants of corn, rice, soybeans, and wheat prices from
2002 to 2009. These commodity prices rose rapidly from 2006 to mid‐2008. They point to demand and
supply explanations for this upward trend. The demand‐side factors include:
 (i) Rising world demand for direct consumption and animal feed: emerging nations can afford a
 more diversified food consumption basket;
 (ii) Ethanol/biofuels significantly increased the demand for a limited supply of commodities;
 (iii) Increased activity in the futures market, also referred to as the financialization of the
 commodity markets. In this process, commodity prices depend not only on its primary
 supply and demand but also on financial variables and investors' behavior in derivative
 markets (Creti, Joëts, and Mignon, 2013).
 Supply‐side factors include:
 (i) Increased oil and fertilizer prices increase costs and limit production;
 (ii) Low R&D investment in agriculture dampens production growth;
 (iii) Trade barriers and export restrictions create scarcity and increase commodity prices;
 (iv) Droughts in large grain‐producing nations lower global production;
 (v) Dollar devaluation: most commodity prices are in USD;
 Cooke and Robles (2009) find that futures markets and proxies for speculation explain changes in
food prices. For their part, Aliaga, Mora‐García and Mulder (2020) finds that speculation activity in the
futures market for the 'C' Coffee Contract lowers price volatility.
 FAO (2009) describes factors affecting agricultural commodities' price formation process in the
exporting country and the world. First, price formation in the exporting region results from the
interaction between domestic production and domestic demand. Factors that determine domestic
production in the exporting countries include input costs, domestic policies, weather, prices, use in
previous years, etc. Domestic demand in exporting countries depends on population, income, taste,
prices of complementary products, prices of competitive products, meat demand, prices of other feeds,
oil prices, and domestic policies. Second, price formation in the world depends on the interaction
ECLAC The main drivers of arabica coffee prices in Latin America 11

between export supply and import demand. Other factors import duties and exchange rates
(see diagram 1).
 Diagram 1
 Factors affecting agricultural commodity prices
 Initial stocks
 Domestic production Input costs
 Policies
 Weather
 Price formation in exporting country Technology
 Duties
 Exchange rates Commodity prices
 World Utilization previous
 Exports export supply year
 +
 Domestic demand exporting country World price formation Initial stocks
 Human consumption Domestic production
 Feed +
 Industrial use
 World
 Final stocks Imports
 import demand

 Population Duties
 Income Exchange rates Population Price formation in importing country
 Taste SPS/NTBs* Income
 Prices of complementary products Taste
 Prices of competitive products Prices of complementary products
 Demand for meat Prices of competitive products Domestic demand importing country
 Prices of the other feeds …etc. Human consumption
 Oil prices Demand for meat Feed
 Oil prices Industrial use
 Domestic prices Prices of other feeds
 Policies Final stocks
 …etc. …etc..
 Technology
 …etc.

Source: Food and Agriculture Organization of the United Nations (FAO) (2009), The State of Agricultural Commodity Markets 2009. Rome: FAO.

 Sachs et al. (2020) study the determinants of coffee prices. Overall, they find that the main drivers
of the ICO Composite price are:
 (i) Brazilian Real (BRL) to US Dollar (USD) exchange rate and the USD to EUR exchange rate;
 (ii) The increase in productivity in Brazil and Vietnam (higher yields);
 (iii) Consolidation in the roaster and retail sectors;
 (iv) High price elasticity of Brazil supply (unused land may return to coffee production under the
 right price);
 (v) Low price elasticity of the rest of the coffee producing countries;
 (vi) Market power in the roast‐retailer segment; and
 (vii) Monopsonic buyer power of coffee.
 They also point out several factors contributing to direct and indirect coffee production costs.
Direct costs include wages and input costs (fertilizers, pesticides, other agrochemicals, and fuels for
machinery). Indirect costs refer to administrative charges (financial cost, administrative labor, legal and
certification costs), cost of planting and renovation (seedling and shading maintenance), and other
expenses related to infrastructure (drying facilities or depreciation rates).
 They also distinguish between influencing factors at the farm, national, and global levels. At the
farm level, these include land area (if the farm is small, medium, or large) and the type of coffee farm (if
ECLAC The main drivers of arabica coffee prices in Latin America 12

it is diversified, organic, or specialty‐focused). At the national level, they include national policies, labor
policies, minimum wage, demographics, and the existence of coffee associations or organizations or
not. At the global level, they refer to the exchange rate, global demand, international organizations,
and international policies related to coffee and labor.
 Finally, they also distinguish between social and environmental externalities. Social externalities
include child labor, gender, rural labor, forced labor, and health. Environmental externalities include
water use, energy use, and land use (see diagram 2).
 Diagram 2
 Conceptual framework for factors contributing to coffee production costs

 Direct Costs Influencing Factors Externalities

 Labor • Year labor Farm Level Social
 • Harvest labor
 • Family labor • Small/medium/large
 • Child labor • Specialty/diversified/off‐farm
 activity
 Inputs • Fertilizers
 • Pesticides
 • Other agro.
 Farm Level
 • Fuels for machinery
 • National policies
 • Labor policies Environmental
 Indirect Costs • Minimun wage
 • Demographics –labor in agricultura • Water use
 Administration • Financial costs • Coffee associations and • Energy use
 • Admin labor • Land use
 organization
 • Legal costs
 • Certification costs
 Global Level
 Planting & • Seeding
 • Shading materials • Exchange rate
 Renovation
 • Demand
 • International organization
 Infrastructure • Facilities (e.g., drying)
 • International policies (coffee,
 labor)

Source: Jeffrey Sachs, Kaitlin Y. Cordes, James Rising, Perrine Toledano, and Nicolas Maennling (2019), “Ensuring Economic Viability and
Sustainability of Coffee Production,” New York: Columbia Center on Sustainable Investment.

 They develop six econometric coffee supply and demand models. Based on projections of
income, population growth, and climate change, these models allow them to evaluate how these
factors will affect the coffee industry in the following 30 years. In particular, they develop:
 (i) Demand model: a two‐staged linear econometric model. The first stage explains the
 international price of coffee ∗ on the basis of the quantity produced in the previous
 growing season, past international price ∗ , and a time trend . The second stage
 explains coffee consumption from the predicted international price from the first stage
 ∗ , current income and a time trend ;
 (ii) Harvest (Yields) model: a linear econometric model that explains domestic yields as a
 function of climatic variables . This model helps to create yield predictions based
 on projections of climate change ;
 (iii) Yield gaps model: captures the relationship between model parameters in the harvest
 model and suitability, as estimated by the Global Agro‐Ecological Zone (GAEZ) project;
 (iv) Planting model: a linear econometric model that explains farmers’ decisions to expand
 planted area Δ from farm prices , yields , and planted area . This model
ECLAC The main drivers of arabica coffee prices in Latin America 13

 facilitates planted area predictions Δ based on a prediction from the yields model 
 and projections of suitable land for cultivating coffee from GAEZ;
 (v) Production model: combines the results from the yields model and the planting model
 to obtain a forecast of domestic coffee production , and then combines the
 domestic coffee production into a global coffee production = ∑ . The combined
 estimated global coffee production feeds into the Demand Model to calculate the next year's
 international prices;
 (vi) Farm gate price model: a linear econometric model that explains farm gate prices from
 international prices ∗ , underlying trends , and records of differentials . This model uses
 the predictions of international prices from the demand model ∗ to get a forecast of farm
 gate prices ̂ ∗ which are used for the planting model because they affect the planted area;
 (vii) Stock model: a linear econometric model that explains stock levels from past behavior
 and the international price ∗ . This model provides predictions of the stock levels
 that influence the equilibrium price.
 In sum, many commodity price movement models (particularly for grains) have employed stocks‐
to‐use ratio, weather variables, international trade, and substitute products (Goodwin et al, 2005).
 Various methods for modeling structural change have been proposed in the literature to estimate
the determinants of commodities prices. Procedures vary from the use of multivariate simple linear
regressions (Westcott and Hoffman, 1999) to models using most complex techniques, such as VAR
models (Deaton and Laroque, 1992) or gradual switching regressions (Goodwin and Piggott, 2001).
Panel regressions are also extensively used in the literature. They are used to minimize endogeneity
problems arising from time variables (Rezitis, 2015; Aghion et al., 2009; Bleaney and Greenaway, 2001;
Cavalcanti, Mohaddes and Raissi, 2011). These studies apply recent findings from econometric theory
(Wooldridge, 2012; Kapetanios et al., 2011; Pesaran, 2006). Following these latter studies, this paper
uses a heterogeneous panel analysis to shed light on the determinants of coffee prices.
ECLAC The main drivers of arabica coffee prices in Latin America 15

 II. A simple pricing model for arabica coffee

Green coffee is a storable agricultural commodity. As a result, stocks play a strategic role.2 Green coffee
producers can build supplies in periods where prices are low, and sell them on the expectation that prices
recover the next year. Following Goodwin, Schenpf and Dohlman (2005), an equilibrium model for a
storable commodity in a competitive market generally consists of a supply equation, a demand
equation, stocks, and an identity describing equilibrium. Thus, a reduced‐form expression for prices will
relate prices to factors that influence supply and demand.
 Supply ( ) is a function of price ( ) (or, more accurately, expected price) and factors ( ) reflecting
production shocks:
 = , 
 Factors that shift the supply curve are those that affect marginal costs and marginal revenues:
 (i) A rise in input prices increases the marginal cost of production and hence shifts the supply curve
 inwards, increasing the equilibrium price. The price of inputs includes wages, price of fertilizers and
 pesticides (including fungicides, herbicides, and insecticides), cost of land (e.g. property taxes),
 transportation costs (gasoline, diesel or oil), interest rate, inflation (affecting the general cost and
 price levels) and exchange rate (which affects the terms of trade).
 (ii) An increase in the price of substitutes in production, such as other commodities, shift
 resources from the production of coffee towards the production of commodities with a higher
 profit margin. This shifts the supply curve inwards and increases the equilibrium price. An
 increase in the price of complementary goods has the opposite effect.
 (iii) A rise in the international coffee price increases national coffee production and shifts the
 supply curve downwards, decreasing the equilibrium price.

2
 Under the right conditions, green coffee can be stored between one and two years without losing quality properties.
ECLAC The main drivers of arabica coffee prices in Latin America 16

 (iv) A productivity improvement, due to changing weather conditions, GDP per capita and yield
 (tons per hectare), shifts the supply curve downwards, decreasing the equilibrium price.
 Market equilibrium requires − − = 0, where are stocks. This allows for a price‐dependent
reduced form expression to be solved that is a function of stocks and supply and demand shifters:
 = , , 
 It is essential to differentiate between different prices . Wholesale and retail market prices
of roasted/soluble coffee are prices paid by consumers to the commercial distributor or outlet. World
prices may be the spot or future price of the Coffee 'C' contracts in USD per 100 pounds of green coffee,
traded at the InterContinental Exchange (ICE). It may also be the ICO Composite prices for Arabica,
Brazil Naturals, Colombia and Other Milds, which are prices for prompt shipment (within 30 calendar
days—as opposed to immediate shipment) and sales from origin. The export price, or free on board
(FOB) price, is the exporting price. The farm gate prices or producer prices are prices received by
farmers at the farm gate, before transportation costs from the farm gate to the nearest market or first
point of sale, excluding warehousing cost, processing cost and market charges for selling the product.
 The particular interest of this work lays on the determinants of the ICO Composite prices and the
farm gate price. Our focus is on Arabica coffees, as we have access to daily ICE and ICO prices, while farm
gate prices are monthly. When we gathered data for each price variable, we tried to be as comprehensive
as possible. The following section describes the definition of the variables data and its sources.
ECLAC The main drivers of arabica coffee prices in Latin America 17

 III. Definition of variables and data sources

This section presents the variables included in our estimated models and describes their data sources.
We first list the supply variables, followed by demand variables and finally prices, our dependent
variable. The frequencies are the main limitation when defining the variables, as some are available only
annually, while we are interested in monthly trends.

 A. Data sources for factors that shift the supply curve
Table 1 list the data sources and definitions of the factors that shift the supply curve. Also, we collected
data on annual green coffee production, imports, and exports from the Production Supply and
Distribution (PSD) of the United States Department of Agriculture (USDA) (annual). We used monthly
data on Colombian coffee production from the FNC.
 As a proxy for transportation costs, we used data on the West Texas Intermediate (WIT) oil spot
price from FRED. As a proxy for the lending interest rate, we used the 3‐Month Treasury Constant Maturity
Rate from FRED; and monthly data on the average lending rate in Colombia (“tasa de cambio
representative del mercado”) from the Bank of the Republic of Colombia (“Banco de la República de
Colombia”). We use the growth rate of the monthly Producer Price Index of All Commodities (PPIACO) of
the United States from FRED, the monthly Consumer Price Index (CPI) from the Banco de la República de
Colombia, and the annual CPI for other countries from the World Development Indicators (WDI) of the
World Bank. We do not use the CPI of the United States because it is correlated with the PPIACO. As a
proxy for the exchange rate, we use the Colombian Peso (COP) to United States Dollar (USD) and Brazilian
Real (BRL) to USD real exchange rates from FRED. Other exchange rates are taken from the WDI.
 We tried to include property taxes, wages, gasoline, and diesel prices. However, we could not find
a historical data set for property taxes for the LAC countries in our sample. We found annual data for
minimum wages from the International Labor Organization, but data were missing for many countries.
We could also not find a good source for gasoline and diesel annual prices for all LAC countries. We also
gathered data for the yearly cost of fertilizers and pesticides (fungicides, herbicides, and insecticides) from
ECLAC The main drivers of arabica coffee prices in Latin America 18

the Food and Agriculture Organization (FAO) as proxies for input prices. However, as these data contained
many missing values, they were excluded to avoid a significant reduction in our country sample.
 As a proxy for productivity, we use weather, GDP per capita and yield. Annual GDP per capita
comes from WDI, and annual yield (in tons per hectare) from FAO. We also use the annual amount of
arable land available from the WDI.
 Following the literature, we define weather as a factor that shifts the supply curve.3 We gather
daily temperature and precipitation data from Climate Engine (https://clim‐engine.appspot.com/
climateEngine), specifically, from the CFS Reanalysis, ERA5 and CHIRPS historical datasets for coffee‐
producing Latin American countries. Climate Engine provides a simple average of temperature and
precipitation at a given scale (computation resolution) that change from one dataset to another. We
then calculate a monthly mean temperature and accumulated precipitation by country.
 Table 1
 Data sources and definitions of factors that shift the supply curve and its data sources
Factor Variable Frequency Source Countries Expected Sign
Input Prices Price of fertilizers Annual FAOSTAT LAC +

 Price of pesticides Annual FAOSTAT LAC +

 Price of WTI Monthly FRED +

 3-Month Treasury Constant Maturity Rate Monthly FRED United States +

 Exchange rate Monthly BRPC Colombia +

 Producer Price Index for All Commodities (PPIACO) Monthly FRED United States +
 Consumer Price Index (CPI) Annual WDI LAC +

 Consumer Price Index (CPI) Monthly BRPC Colombia +

 COP to USD Monthly FRED +

 BRL to USD Monthly FRED +

 National currency exchange rate Annual WDI LAC +

Substitutes & PPIACO Monthly FRED United States + (substitute)
Complements - (complement)

International Spot Price Daily ICE +
Price
 Composite Prices Daily ICO +

Factors Rain Daily Climate LAC -
affecting Engine
Productivity Temperature Daily Climate LAC -
 Engine
 GDP per capita Annual WDI LAC -

 Arable Land Annual WDI LAC -

 Yield Annual PSD LAC -

Source: Elaboration by the authors.

 A potential weakness of our work is that we use a simple average of our climatic variables, which
means that all observations across a given grid get the same weight within a country. This method fails
to consider that some areas are more suitable for cultivating coffee than others. Further research would
ideally use a weighted average by the suitability for growing coffee in each cell of the grid, using some
services such as the Global Agro‐ecological Zones (GAEZ) services.4

3
 However, we also want to note that weather may also affect the demand for coffee (consumption) through seasonality with coffee
 consumption peaking during the colder months of the year and declining during the warmer months (http://www.futuresmag.
 com/2017/03/19/exploiting‐coffee‐seasonality).
4
 See http://www.gaez.iiasa.ac.at/w/ctrl?_flow=Vwr&_view=Welcome&idAS=0&idFS=0&fieldmain=main_&idPS=0.
ECLAC The main drivers of arabica coffee prices in Latin America 19

 B. Data sources for factors that shift the demand curve
Table 2 lists the data sources and definitions of the factors that shift the demand curve. Also, we
collected annual green coffee consumption data from PSD of USDA.
 (i) As proxy for demographic size, we used data on annual population from the WDI.
 (ii) As proxy for income, we collected data on annual gross national income and GDP per capita
 from the WDI.
 (iii) As a proxy for changes in the prices of substitutes and complements in coffee consumption,
 we use the growth rate of the monthly Producer Price Index of All Commodities (PPIACO) of
 the United States from FRED.
 Table 2
 Data sources and definitions of factors that shift the demand curve and its data sources
 Factor Variable Frequency Source Countries Expected Sign
 Demographic Factors Population Size Annual WDI LAC +
 Income Gross National Income (GNI) Annual WDI LAC +
 GDP per capita Annual WDI LAC +
 Gini Index Annual WDI LAC +
 Price of Related Commodities PPIACO Monthly FRED United States + (substitutes)
 - (complements)

Source: Elaboration by the authors.

 C. Stocks
We use annual end‐of‐year stocks of green coffee from PSD. From ICE we gather historical end of month
certified Coffee C stocks by port held by ICE Futures U.S. Licensed Warehouses measured in total bags.

 D. Prices
Price data comes from different sources. First, we obtain daily spot prices on the US 'C' Contract for
Arabica Coffee traded at the InterContinental Exchange (ICE) market, a continuous contract based on
Arabica coffee.5
 The above data are complemented with daily ICO indicator prices for Colombian Milds, Other
Milds and Brazilian Naturals that belong to the Arabicas group. ICO indicator prices are a weighted
average of daily prices of green coffee on the physical markets in the United States, France and
Germany. Prices are for prompt shipment (within 30 calendar days as opposed to immediate shipment)
and sales from the origin, in USD per 100 pounds of green coffee. These quotations come from at least
five traders and brokers. So ICO's prices are spot prices. There is some relationship between ICO
indicator prices and the Coffee C Contract traded at ICE, because traders and brokers often trade green
coffee as Coffee C Contract plus a margin. We then take a weighted average of these three, the weights
are from ICO (2011) —Colombian Milds 12%, Other Milds 23% and Brazilian Naturals 31%— and
construct one composite category for Arabica coffee. Appendix Table A4 show the classification of

5
 The market and exchange names of the Coffee C contract has changed over time. Until August 2007 it was the Coffee, Sugar and
 Cocoa Exchange. From September 2007 until December 2019 it was the New York Board of Trade. From January 2020, it is the ICE
 Futures U.S. The price data is taken from Barchart [online] www.barchart.com.
ECLAC The main drivers of arabica coffee prices in Latin America 20

exporter countries by type of coffee according to ICO. When working with ICE and ICO prices, we can
take averages of daily prices at weekly, monthly and annual frequencies.
 We also collected monthly prices paid to Colombian coffee growers from the National Federation
of Coffee Growers of Colombia (Federación Nacional de Cafeteros de Colombia, FNC). This data is in
Colombian Pesos (COP) per 125 kilos of dry parchment coffee (a unit known in Colombia as 'carga'), and
we apply the COP to USD exchange rate to transform it into USD prices.
 Finally, annual farm gate prices from ICO, in USD per 100 pounds of green coffee, for many
coffee‐growing countries, allow us to use a panel data econometrics.
 Table 3
 Data sources and definitions of stocks prices
 Expected
 Variable Frequency Source Countries
 Sign
 Stocks End-of-year stocks Annual PSD LAC -
 Coffee C stocks Monthly ICE -
 Prices US 'C' Contract Spot Price Daily ICE
 Brazilian Naturals Monthly ICO
 Colombian Milds Monthly ICO
 Other Milds Monthly ICO
 Prices paid to coffee growers Monthly FNC
 Farm gate prices Annual ICO LAC

Source: Elaboration by the authors.

 E. Data description
Figure 1 plots the spot prices for the Coffee 'C' Contract traded at the InterContinental Exchange (ICE),
and the ICO composite prices for Colombia Milds, Brazil Naturals, and Other Milds. Prices for Colombian
Milds closely follow the spot prices at ICE, while prices for Brazil Naturals and Other Milds have
historically been under the spot price.
 Figure 1
 Arabica coffees: weekly spot prices for coffee ‘C’ contract traded at ICE and ICO composite prices for the three
 categories (Brazil, Colombia and other milds), 1997‐2020
 300

 300
 Price (USD $ per 100 lb)
 200

 200
 100

 100
 0

 0

 1997w1 2000w1 2005w1 2010w1 2015w1 2020w35

 Colombia Brazil
 Others Spot/Cash

Source: Elaboration by the authors based on sources listed in section 3.
ECLAC The main drivers of arabica coffee prices in Latin America 21

 In 2019, Brazil produced 3.5 times more (Arabica) coffee than Colombia. The latter country made
6.7 times more (Arabica) coffee than the average Latin American and Caribbean nation. Exports follow
a similar pattern (see Appendix Figure A1).
 Figure 2 plots the price differential between the three categories of the ICO composite prices and
spot prices. The price differential for Colombia fluctuates almost randomly around zero and has the
typical behavior of a white noise error, while the price differential for Brazil is typically negative. The
price differential for Other Milds has been negative for most of the time, except for a few weeks around
the year 2015. It is also interesting to note that price differentials for Brazil and Other Mild Arabicas
negatively correlate with the spot price level, meaning that price differentials decrease when prices
increase. Moreover, price differentials for Brazil and Other Mild were smaller when coffee prices were
at their maximum level during the period under study.
 Figure 2
 Arabica coffees: weekly price differentials between Coffee ‘C’ Contract spot prices and ICO Composite
 prices for the three categories, 1997‐2020
 50

 50
 Price premium (USD $ per 100 lb)
 0

 0
 -50

 -50
 -100

 -100
 -150

 -150
 1997w1 2000w1 2005w1 2010w1 2015w1 2020w35

 Colombia Brazil
 Others

Source: Elaboration by the authors based on data sources listed in section 3.

 Another important factor are yields. Many factors determine yields, such as growing techniques,
topography, R&D in the development of new more productive and resistant coffee varieties,
climatological variables that influence the suitability of land for growing coffee. An improvement in
yield shifts the supply curve downwards, decreasing the equilibrium price. Table 4 shows average green
coffee yields (hg/ha) by LAC countries in our sample, from 1961 to 2018. The second column shows the
average yield during the 60s, and the other columns shows the average growth rates of yields in the
following decades.
 The table provides evidence on the significant productivity gains Brazil has experienced since
2000. This is important since Brazil is one of the biggest coffee producers worldwide. The increase in the
Brazilian yield affects the marginal cost and production decisions of coffee farmers. The same amount
of land can produce more green coffee, making it more profitable. Since Brazil is one of the biggest
coffee producers worldwide, changes in its production capacity affect international spot prices.
According to Sachs et al. (2019) this is one of the main factors driving global green coffee prices.
ECLAC The main drivers of arabica coffee prices in Latin America 22

 Table 4
 Decade average of green coffee yields in coffee growing Latin American countries, 1961‐2018

 (hg/ha) Variation (percentage)

 1961-1969 1970-1979 1980-1989 1990-1999 2000-2009 2010-2018

 Brazil 4 617 15.6 10.1 4.4 63.2 45.4
 Colombia 5 784 5.2 20.2 18.6 3.1 -4.0
 Costa Rica 7 592 33.6 38.0 0.9 -14.2 -14.5
 Cuba 1 977 -16.2 22.3 16.3 27.2 6.5
 Dominican Republic 2 969 9.6 36.8 -27.2 -11.2 -34.9
 Ecuador 3 052 4.5 -10.9 25.4 -44.7 -20.5
 El Salvador 9 701 5.8 -12.7 -3.4 -32.5 -25.1
 Guatemala 4 689 28.6 29.3 11.8 15.9 -7.9
 Haiti 5 292 -2.0 -3.0 -0.8 4.0 -0.3
 Honduras 3 560 24.9 41.7 21.0 16.1 16.5
 Jamaica 3 062 -19.6 31.0 38.4 70.5 10.1
 Mexico 4 893 14.2 -6.4 -4.0 -21.6 -24.4
 Nicaragua 3 403 60.7 10.0 5.3 5.8 27.6
 Panama 2 366 -9.3 83.1 8.0 6.7 -1.3
 Paraguay 9 678 -26.2 39.0 -17.0 15.6 37.7
 Peru 5 137 12.6 -5.1 10.9 20.9 2.4
 Bolivarian Republic
 of Venezuela 1 817 10.8 20.9 27.8 14.7 -1.9
 Puerto Rico 2 281 -0.7 28.5 36.6 -11.4 60.6

Source: Elaboration by the authors based on data sources listed in section 3.

 While Honduras has also experienced a constant increase in its yield, it is lower than Brazil's. Most
other countries have experienced a decrease in their yield, for example Costa Rica, which experienced
its highest yield growth during the 1970s. Figure 3 visualizes table 4, highlighting the cases of Brazil and
Costa Rica. Nowadays Brazil is the LAC country with the highest yield, followed by Paraguay, Honduras,
Nicaragua and Costa Rica.
ECLAC The main drivers of arabica coffee prices in Latin America 23

 Figure 3
 Annual yield (hg/ha) of green coffee in Latin American countries, 1961‐2018

 A. Brazil

 25 000

 20 000
 Green coffee annual yield (hg/ha)

 15 000

 10 000

 5 000

 0
 1961
 1963
 1965
 1967
 1969
 1971
 1973
 1975
 1977
 1979
 1981
 1983
 1985
 1987
 1989
 1991
 1993
 1995
 1997
 1999
 2001
 2003
 2005
 2007
 2009
 2011
 2013
 2015
 2017
 B. Costa Rica

 25 000

 20 000
 Green coffee annual yield (hg/ha)

 15 000

 10 000

 5 000

 0
 1961 1968 1975 1982 1989 1996 2003 2010 2017

Source: Elaboration by the authors based on data sources listed in section 3.
ECLAC The main drivers of arabica coffee prices in Latin America 24

 The real exchange rate between the Brazilian Real and the US Dollar is another important factor
that drives fluctuations in the spot price of green coffee. Figure 4 shows the monthly relationship
between the real exchange rate of the Brazilian Real to the US Dollar and the real spot price from
January 1994 to June 2020. Both series appear strongly related to each other (see figure 4). As explained
by Sachs et al. (2019, pg. 14) "a weak Real is positively correlated with higher Brazilian coffee production
and exports and consequently lower coffee prices in US dollars".
 Figure 4
Monthly relationship between the real exchange rate of the Brazilian Real to the US Dollar and the real spot price,
 Jan 1994‐Jun 2020
 250
 200
 Real Value
 150
 100
 50

 1994m1 1999m1 2004m1 2009m1 2014m1 2020m6
 Monthly Date

 BRL/USD Real Exchange Rate Real Spot Price

Source: Elaboration by the authors based on data sources listed in section 3.
Notes: The real spot price comes from dividing the nominal spot price in dollars by the producer price index of all commodities (PPIACO)
and multiplying this by 100.

 The previous results show correlations between time series. However, these relations simply
result from the cointegration of the time series. The next section explains how we control for
cointegration and explains the methodological approach we take to quantify the relationship between
the different factors that drive price levels.
ECLAC The main drivers of arabica coffee prices in Latin America 25

 IV. Methods and results

The three different prices (ICE spot prices, ICO Composite prices, and FAO farm gate prices) are
available at a different frequency (daily, weekly, monthly and annually). Each price frequency has its
own determinants. For example, the effect of weather on prices can be analyzed for annual prices but
not for monthly or weekly prices, except if we know how climatic realizations affect price formation over
short time horizons. By studying annual prices, we can take advantage of a non‐parametric specification
that is flexible enough to allow for each month of the year to affect the annual price differently. This
allows us to know which month's climatic conditions are more relevant.
 Below four different models are considered: the first studies monthly ICO prices (Brazil, Colombia
and Others). The second is a Colombia‐specific model that analyzes monthly farm gate prices on one
hand and annual farm gate prices on the other. The third model uses a fixed effects model and annual
farm gate prices from ICO. Each model considers different determinants. Next, we explain the
regression equation behind each model.

 A. Model 1: ICO composite prices (monthly, Jan‐1997 to Sep‐2020)
This model fits the following equation by ordinary least squares (OLS),
 Δln ∗ = + Δ ln + 3 + Δ ln + Δ ln + Δ ln 
 + ∗ + , (1)
 where Δln( ) = ln( ) − ln( ) is the percentage change of any variable . The dependent
variables in this model are the monthly percentage change in four different international prices ∗ from
ICO composite prices (in USD/100lb of green coffee). We use the price of Brazilian Naturals, Colombian
Milds, Others Milds, and our "Arabica" category—constructed as a weighted average of the former
three. We add a comprehensive set of covariates to the model. Specifically, we include as control
variables the total end of month certified Coffee 'C' stocks, 3 the 3‐Month Treasury
Constant Maturity Rate (a proxy for the international rate), the West Texas Intermediate oil price,
ECLAC The main drivers of arabica coffee prices in Latin America 26

and ∗ the monthly growth rate of the Producer Price Index of All Commodities (inflation). In addition
to the the BRL‐USD real exchange rate we also included the the COP‐USD real exchange
rate. Both the BRL/USD and COP/USD have a very similar behavior and we wanted to test where the
effect was coming from. All variables are measured monthly and we correct standard errors by using a
robust covariance matrix (Eicker‐Hubert‐White).
 Table 5 show the results from equation (1). The first column shows results for the weighted average
of all ICO Arabica price categories; the second column focuses on Brazil Naturals; the third on Colombian
Mild; and the last column on Other Milds. We find that (i) the growth in the BRL/USD real exchange rate
has a positive and statistically significant effect at the 1% level on the growth rate of all four ICO prices,
but this effect is bigger for Brazil Naturals. This means that an increase (depreciation) in the BRL/USD real
exchange rate increases all Arabica ICO prices. (ii) regardless of the ICO price considered, the rate of
growth of end‐of‐month stocks both have a negative and statistically significant effect at the 5% to
10% level on the growth rate of all four ICO prices considered. This means that stocks acts as a buffer to
price variations, an increase in stocks decreases prices. (ii) we do not find statistical evidence that the
growth in the COP/USD real exchange rate affects the growth rate of any of the Arabica ICO prices, as the
share of Colombia in coffee sales is not sufficiently large to affect coffee prices. (iii) Inflation in commodity
prices also affects prices positively and has a statistically significant effect at the 5% to 10% level on the
growth rate of Arabica, Brazil and Others, but it does not affect the Colombia ICO prices. (iv) Neither the
interest rate nor the oil price have any effect on any of the ICO prices.
 Table 5
 Determinants of monthly ICO prices from Arabica coffee (our Arabica category, Brazilian Naturals,
 Colombian Naturals, and Other Milds) using Model 1
 (1) *** (2) *** (3) *** (4) ***
 Arabica*** Brazil*** Colombia*** Others***
 3 -0.2098 -0.2333 -0.1839 -0.1891

 (0.2478) (0.2403) (0.2778) (0.2311)
 Δ ln( ) 0.1032 0.0325 0.1467 0.1653
 (0.1553) (0.1682) (0.1532) (0.1475)
 Δ ln( ) -0.0600 -0.0785 -0.0306 -0.0682
 (0.0539) (0.0573) (0.0559) (0.0507)
 ∗ 0.8520* 0.9976** 0.6349 0.8473**
 (0.4391) (0.4787) (0.4311) (0.4221)
 - -
 -0.0470*
 Δ ln( ) -0.0603*** 0.0687*** 0.0632***
 (0.0221) (0.0210) (0.0258) (0.0239)
 Δ ln( ) 0.4420*** 0.5690*** 0.3523*** 0.3668***
 (0.1165) (0.1260) (0.1108) (0.1148)
 Constant 0.3439 0.2947 0.3462 0.3816
 (0.5165) (0.5502) (0.5269) (0.4965)
 N 274 274 274 274
 0.09 0.12 0.06 0.09
 Adjusted 0.07 0.10 0.04 0.07
 F-statistic 5.55 7.74 3.58 4.95
 AIC 1 829.13 1 842.74 1 859.28 1 797.14
 BIC 1 854.43 1 868.03 1 884.57 1 822.43
 p-JB 0.00 0.00 0.00 0.00
 Durbin-Watson d 1.77 1.69 1.88 1.67

Source: Elaboration by the authors based on data sources listed in section III.
Notes: This table shows coefficient estimations using Model 1 and equation (1). The first column show results when using as dependent
variable the monthly growth rate in our ICO Arabica composite price, the second uses the monthly growth rate in the ICO Brazilian Naturals
price, the third uses the monthly growth rate in ICO Colombian Milds price, and the fourth uses the monthly growth rate in Other Milds price.
The row called p‐value Normal is the p‐value for the D’Agostino, Belanger, and D’Agostino (1990) Normal Distribution test of the residuals.
Robust standard errors, i.e. Huber (1967) and White (1980, 1982), are in parentheses. Significant at * 10%, ** 5%, *** 1%.
ECLAC The main drivers of arabica coffee prices in Latin America 27

 Specification Checks. The last rows in Table 5 shows the Durbin‐Watson d statistics for first order
autocorrelation that range from 1.67 to 1.88; it also shows the p‐value for the D'Agostino, Belanger, and
D'Agostino (1990) test and provides evidence against a Normal Distribution of the residuals. Table A2
show variance inflation factors of multicollinearity, that are lower than 1.77 in all cases. It also shows the
Dickey‐Fuller statistic for the presence of a unit root in each of the variables (both dependent and
independent) in the model. The results show that we can reject the null hypothesis of a unit root at all
common significance levels, except for the international interest rate. The R^2 range from 0.06 to 0.12,
but increases to 0.77 to 0.83 if we control for the spot prices (results are available upon request to the
authors). The F‐statistics range from 4.95 to 7.74 and are highly significant.

 B. Model 2: colombian prices and production
 (monthly, Jan‐1997 to Jun‐2020)
This model attempts to bring forward the most important factors correlated to the Colombian coffee
prices. We take advantage that the ICO composite price for Colombian Milds consists mostly of coffee
grown in Colombia (although Kenya and Tanzania also trade coffee under this category). Moreover, we
have data on prices paid to coffee growers from FNC. Therefore, this model attempts to quantify if
national shocks occurring in Colombia affect (a) prices at warehouses outside of Colombia and (b) the
pricing process by the FNC. We fit the following equation by ordinary least squares (OLS),
 Δln( ) = + Δ ln( ) + 3 + Δ ln( ) + Δ ln( ) + Δ ln( )
 + ∗ + + + + + Δln ( ∗ ) + , (2)
 where the dependent variable in this model is either the Colombian Mild ICO Composite price,
the price paid to coffee growers in Colombia, or the Colombian coffee production in month . We
dollarize the prices paid to Colombian coffee growers by dividing the price reported by the FNC by the
COP/USD real exchange rate.
 Equation (2) includes the same covariates from equation (1), in addition to the Colombian
interest rate and the growth rate in the Colombian CPI (inflation). We also add the average
cumulative rain and the average temperature in the Colombian territory. When the outcome of
interest is the price paid to Colombian coffee growers or the Colombian production, we included ∗ the
ICO Composite Colombian Mild international price. In this context, we use the variable as a proxy
to represents transportation costs and other processing activities that requires the use of gasoline or
diesel. All variables have a monthly frequency. Moreover, we correct standard errors by using a robust
covariance matrix (Eicker‐Hubert‐White).
 Table 6 presents results from equation (2). The first column shows results for the ICO Colombian
Mild prices, the second for the price paid to coffee growers from FNC, and the third for FNC coffee
production. The results in column (1) confirm the results from column (3) of Table 5. We also find that
(i) there is no statistically significant effect of national shocks (except for the interest rate) on Colombia
Mild ICO prices. Regarding column (2), we find that (ii) the Colombian Milds international price has a
positive and statistically significant effect at the 1% level on the price that producers receive when they
sell their coffee to the FNC. This means that there is a certain degree of pass‐through from international
markets to national markets, that is below unity (a pass‐through of 79.4% of the Colombian Mild price
in a given month). A future study could examine how this this pass‐through has evolved over time.
You can also read